Escolar Documentos
Profissional Documentos
Cultura Documentos
Introduction
This is the pdf-version of the Number Theory Formulary on MathLinks/ArtOfProblemSolving (http://www.mathlinks.ro/Forum/viewtopic.php?t=76610). All contributors are welcome to post new theorems at http://www.mathlinks.ro/Forum/viewtopic.php?t=76609. Contributers till now: {x}, Myth, pbornzstein, Schoppenhauer, mathmanman, Xixas, campos, t0rajir0u, bodan, dule 00.
CONTENTS
CONTENTS
Contents
1 Symbols and conventions 1.1 1.2 Sets of numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Denitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2.1 1.2.2 1.2.3 General stu . . . . . . . . . . . . . . . . . . . . . . . . . . . . Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Counting function and densities . . . . . . . . . . . . . . . . . . 4 4 4 4 5 6 7 8 9 9 13 14 15 17 18
2 Elementary congruences and divisors 3 Identities 4 Floor function 5 Number theoretic sums 6 Arithmetic functions 7 Sums of squares 8 p-adic numbers, Hasse-Minkowski 9 Legendres and Jacobis symbols, quadratic reciprocity law 10 Representations
CONTENTS
CONTENTS
11 p-adic valuations 12 Primes 13 Additive properties 14 Multiplicative functions 15 Irreducibility of polynomials 16 Finite dierences
19 19 20 21 22 22
1
1.1
Z: the integers (a unique factorisation domain). N: the positive integers, meaning those > 0. P: the positive primes. Q: the rationals (a eld). R: the reals (a eld). C: the complex numbers (a algebraically closed and complete eld). Qp : the p-adic numbers (a complete eld); also Q0 := Q and Q := R is used sometimes. Zn = Z/nZ: the residues mod n (a ring; a eld for n prime). When M is one of the sets from above, then M + denotes the numbers > 0 (when dened), analogous for M . The meaning of M will depend on M : for most cases it denotes the invertible elements, but for Z it means the nonzero integers (note that this denitions coincide in most cases). + A zero in the index, like in M0 , tells us that 0 is also included.
1.2
1.2.1
Denitions
General stu
For a set M , |M | = #M denotes the number of elements of M . . a divides b (both integers) is written as a|b or sometimes as b. . a. Then for m, n Z, gcd(m, n) or (m, n) is their greatest common divisor, the greatest d Z with d|m and d|n (gcd(0, 0) is dened as 0) and lcm(m, n) or [m, n] denotes their least common multiple, the smallest non-negative integer d such that m|d and n|d. When gcd(m, n) = 1, one often says that m, n are called coprime. For n Z to be squarefree means that there is no integer k > 1 with k 2 |n. Equivalently, this means that no prime factor occurs more than once in the decomposition. 4
1.2 Denitions
Factorial of n: n! := n (n 1) (n 2) ... 3 2 1. n! k Binomial coecients: n = k!(n = Cn . k k)! For two functions f, g : N C the Dirichlet convolution f g is dened as . f g (n) := d|n f (d)g n d A (weak) multiplicative function f : N C is one such that f (a b) = f (a) f (b) for all a, b N with gcd(a, b) = 1. Some special types of such functions: Eulers totient function: (n) = (n) := |{k N : k n, gcd(k, n)}| = |Z n |. Moebius function: 0 i n is not squarefree (n) := . (1)s where s is the number of prime factors of n otherwise Sum of powers of divisors: k (n) := d|n dk ; often is used for 0 , the number of divisors, and simply for 1 . For any k, n N it denotes rk (n) := {(a1 , a2 , ..., ak ) Zk | representations of n as sum of k squares. a2 i = n} the number of
Let a, n be coprime integers. Then ordn (a), the order of a mod n is the smallest k N with ak 1 mod n. For n Z and p P, the p-adic valuation vp (n) can be dened as the multiplicity Q , m, n Z by of p in the factorisation of n, and can be extended for m n m vp n = vp (m) vp (n). Additionally often vp (0) = is used. For any function f we dene (f )(x) := f (x + 1) f (x) as the (upper) nite dierence of f . Then we set 0 (f )(x) := f (x) and then iteratively n (f )(x) := (n1 (f ))(x) for all integers n 1.
1.2.2
Symbols
Legendre symbol: for a Z and odd p P we dene when x2 a mod p has a solution x Z p 1 a := 0 i p | a p 1 when x2 a mod p has no solution x Zp i Then the Jacobi symbol for a Z and odd n = pv i (prime factorisation of n) is dened as:
a n
a pi
vi
. 5
1.2 Denitions
Hilbert symbol: let v P {0, } and a, b Q v . Then (a, b)v := 1 1 i x2 = ay 2 + bz 2 has a nontrivial solution (x, y, z ) Q3 v otherwise
is the Hilbert symbol of a, b in respect to v (nontrivial means here that not all numbers are 0).
1.2.3
When A N, then we can dene a counting function a(n) := |{a A|a n}. One special case of a counting function is the one that belongs to the primes P, which is often called . With counting functions, some types of densities can be dened:
n) Lower asymptotic density: L d(A) := lim inf n a(n n) Upper asymptotic density: U d(A) := lim supn a(n n) Asymptotic density (does not always exist): d(A) := limn a(n n) Shnirelmans density: (A) := inf n a(n P
Dirichlets density(does not always exist): (A) := lims1+0 PaA as aN L d(A) and U d(A) are equal i the asymptotic density d(A) exists and all three are equal then and equal to Dirichlets density.
as
Often, density is meant in relation to some other set B (often the primes). Then we need A B N with counting functions a, b and simply change n into b(n) and N into B : Lower asymptotic density: L dB (A) := lim inf n Upper asymptotic density:
U dB (A) a(n) b(n) (n) lim supn a b(n) a(n) b(n) as s aB a
aA
:=
Asymptotic density (does not always exist): dB (A) := limn Shnirelmans density: B (A) := inf n
a(n) b(n)
Dirichlets density(does not always exist): B (A) := lims1+0 Again the same relations as above hold.
P P
Gauss theorem : If a|bc and gcd(a, b) = 1, then a|c. The Gauss theorem comes from : Bezouts identity : The set {ax + by |x, y Z} is the set of all the multiples of gcd(a, b), that is to say : aZ + bZ = gcd(a, b)Z Fermats little theorem: For any positive integer a and every prime p it is ap a mod p. Generalization: Theorem of Euler-Fermat: If gcd(a, m) = 1 then a(m) 1 mod m. Wilsons theorem: For prime p it is (p 1)! 1 mod p. Polynomial congruences: For any polynom f with integral coecients and any integers a, b with a b mod m for some integer m it is f (a) f (b) mod m. Lucas theorem: a ai k mod p where ai s and bi s are the digits of a and b expressed in base p i=0 bi b (p is a prime) with leading zeros allowed. Wolstenholmes Theorem (number 1): 2p 2 mod p3 for p P 5 p Wolstenholmes Theorem (number 2): 1 = 1+ 1 +1 + ... + p with (m, n) = 1 and p is a prime greater than or equal to Let m n 2 3 1 2 5. Then p divides m.
3 IDENTITIES
Identities
Identity of Sophie Germain: For all integers a, b it is a4 + 4b4 = (a2 + 2b2 + 2ab)(a2 + 2b2 2ab). Sum-of-n-squares-identities: - Two squares: (a2 + b2 )(c2 + b2 ) = (ac bd)2 + (ad + bc)2 - Four squares: (a2 + b2 + c2 + d2 )(e2 + f 2 + g 2 + h2 ) = (ae bf cg dh)2 + (af + be + ch dg )2 + (ag + ce + df bh)2 + (ah + de + bg cf )2 - Eight squares: (a2 + b2 + c2 + d2 + e2 + f 2 + g 2 + h2 )(m2 + n2 + o2 + p2 + q 2 + r2 + s2 + t2 ) = 2 2 2 2 2 2 2 u2 1 + u2 + u3 + u 4 + u 5 + u6 + u7 + u 8 where u1 = am bn co dp eq f r gs ht u2 = bm + an + do cp + f q er hs + gt u3 = cm dn + ao + bp + gq + hr es f t u4 = dm + cn bo + ap + hq gr + f s et u5 = em f n go hp + aq + br + cs + dt u6 = f m + en ho + gp bq + ar ds + ct u7 = gm + hn + eo f p cq + dr + as bt u8 = hm gn + f o + ep dq cr + bs + at (see also http://www.geocities.com/titus piezas/DegenGraves1.htm ) Similar to the previous ones: (a2 + nb2 )(c2 + nd2 ) = (ac nbd)2 + n(ad + bc)2 Theorem: (Leibnitz): (x1 + x2 + + xm )n =
k1 ,...,km >0 k1 ++km =n
n m x k1 x k2 x k m . k1 , k2 , . . . , km 1 2
ak c k
k=1 k=1
bk dk
k=1
ak dk
k=1
bk c k
=
1i<j n
(ai bj aj bi ) (ci dj cj di ) .
Vandermondes identity: 8
max{k,n}
=
l=0
m kl
n l
Vn (a1 , a2 , . . . , an ) =
1 1 a1 an 1 1 1 a2 an 2 . . . . . . . . . . . . n1 1 an an
Floor function
On dealing with the oor function: 1. Let n, m N, then m m mod n = m n n Remark: Perhaps this could work with n, m R but who would use it ? 2. Let m N, n Z, x R, then
m1
k=0
m1 (m, n) m nk + x x + n+ = (m, n) m d 2 2
m1
3. Let m N, x R, then mx =
x+
k=0
k m
1. Let n N
(j ) =
j =1 n
3 2 n + O (n log n) 2
(j ) =
j =1
2. Let n N
d (j ) = n log n + (2 1)n + O
j =1
3. Let n, k N
k (j ) =
j =1
1 k+1
1 j 1+k
n1+k + Rk (n)
j =1
where
O(n), when 0 < k < 1 O(n log n), when k = 1 Rk (n) = O(nk ), when k > 1 2. Let Q(n) denote the number of squarefree integers less than n. Then
n
4. Let n
Q(n) =
j =1
2 (j ) =
6 2 n + O n 2
S=
n=1
f (n)
f pk
p k=0
=
n=1
f (n)
f (n) =
n=1 p
1 1 f (p)
(1 f (p))
2 (d)f (d) =
d|n p|n
(1 + f (p))
1, if n = 1 0, if n > 1
9. Let n N, then
j =1
1 = C1 log n + C2 + O (j )
log n n
(j ) = n log log n + Bn + O
j =1 n
n log n n log n
(j ) = n log log n + (B + C )n + O
j =1 n
1 = log log x + B + O p
1 log x
k2,pk x
where p runs through primes and C is a constant. 12. Let n N, then r2 (n) = x + O x1/3 log x
nx
Let n
pn
z p
1 log n
1 d
n 2(n)
15. Let k, l be two positive integers with (k, l) = 1, then 1 1 = log log x + O(1) p (k )
f (p) | p
Cn
pk n
|f pk |2 pk
32).
17. Let f be a strongly additive function, and n a positive integer. Then |f (m)
mn pn
f (p) | p 12
2Cn
pn
|f (p)|2 p
6 ARITHMETIC FUNCTIONS
where p runs through primes and C is a constant (C Some other sums 1. Abelian summation
32).
ai b i =
i=1 i=1
ai b n
m=1 i=1
ai (bm+1 bm )
n 2. Let (am )n m=1 , (bm )m=1 be two nite sequence of real numbers. Then n n n n k 1
ak
k=1
k=1
bk
=n
k=1
ak b k
k=2 j =1
(ak aj ) (bk bj )
Or equivalently
n n n
ak
k=1
k=1
bk
=n
k=1
ak b k
(ak aj ) (bk bj )
1j<kn
j =1
1 1 = ln n + + +O j 2n
1 n2
Arithmetic functions
1 p n d
=
d|n
(d)
2 (d) (d)
13
7 SUMS OF SQUARES
4. Let n N, then
5. Let n
for some positive constant c > 0. 6. For all composite numbers n it holds (n) n n
7. Let pn be the n-th prime number, then an log n for two constants 0 < a < b. 8. Let n be a positive integer, then (n) 9. Let n be a positive integer, then d(n) 2 n lg2 n pn bn log n
Sums of squares
2) Sum of two squares: A positive integer n can be represented as sum of two perfect squares i all prime factors p 3 mod 4 of n occur an even number of times in the factorisation of n. n can be written as sum of squares = 0 i the previous condition holds and it has at least one prime factor 1 mod 4 or v2 (n) is odd. There are exactly r2 (n) = 4
dN d|n d1 mod 2
(1)
d 1 2
=4
pP p1 mod 4
(vp (n) + 1)
14
8 P-ADIC NUMBERS, HASSE-MINKOWSKI dierent solutions (a, b) Z2 to n = a2 + b2 . 3) Sum of three squares: Write n as n = 4k u, k, u N0 with 4 u (but u can be even). Then n can be written as sum of three squares i u 7 mod 8. 4) Sum of four squares: Every positive interger can be written as sum of four squares, and there are r4 (n) = 8
dN 4 d|n
d=
dierent solutions (a, b, c, d) Z4 to n = a2 + b2 + c2 + d2 . 5) Sum of ve squares: As corollary to 4) every integer can be written as sum of ve squares, but there is one more thing to say: except of some small numbers (all < 100), every positive integer can be written as sum of ve nonzero perfect squares. 8) Sum of eight squares: There are r8 (n) = 16
dN d|n
(1)nd d3
p-adic numbers The p-adic integers (for that and only that post written by Zp ) are isomorphic (or by denition identical) to: k a) the (formal) series k=0 ak p with ak {0, 1, 2, ..., p 1}. b) the cauchy-sequences (bk )kN0 of integers in respect to the p-adic valuation ||p = pvp () . c) the projective limit limn Z/pn Z. The last one gives that a polynomial equation p(x) = p(x1 , x2 , ..., xn ) has a solution in Zp i it has one mod any power of p. 15
The p-adic numbers Qp are isomorphic (or by denition identical) to: k a) the (formal) series k=s ak p with ak {0, 1, 2, ..., p 1}. b) the rational cauchy-sequences (bk )kN0 in respect to the p-adic valuation ||p = pvp () . c) the eld of quotients of Zp . Some properties of the Hilbert symbol (holding for any v P {0, } and a, b, c Q v ): - (a, b)v = (b, a)v - (a, 1)v = 1 = (1, b)v - (a, bc2 )v = (a, b)v = (ac2 , b)v - (a, bc)v = (a, b)v (a, c)v Product formula for the Hilbert symbols: Let a, b be rational. Then (a, b)v = 1 for all but netely many v P {} and: (a, b)v = 1
v P{}
Approximation of the Hilbert Symbols: Let a nite set {a1 , a2 , ..., ak } of rational numbers and then for all j K := {1, 2, ..., k } and v P {} an ej,v {1} be given such that: - all but netely many ej,v are equal to 1 - for any j K it holds vP{} ej,v = 1 - there is an xv Q v such that (aj,v , xv )v for all j K Then there exists a rational number x with (aj,v , x)v = ej,v for all (j, v ). The theorem of Hasse-Minkowski: Let f (x) = f (x1 , x2 , ..., xn ) = 0 be any homogenous polynomial equation of degree 2 (so f is a polynomial where every single monomial has degree 2). Then there exists a nontrivial (not all numbers = 0) rational solution x Qn to f (x) = 0 i this equation has a nontrivial solution x Qn v for all v P {}. Corollary: when f has also integer coecients, the equation f (x) = has a nontrivial integral solution i it has a solution mod any integer (where by the Chinese Remainder Theorem we can restrict to perfect powers of primes).
16
Basic facts on the Legendres and Jacobis symbols. The quadratic reciprocity law. Theorem 1. If a b (mod p), then
a p
b p
a2 p
= +1.
Theorem 3 (Eulers criteria). p1 (mod p). a 2 a p Theorem 4. 1 p +1, p = 1 (mod 4), 1, p = 3 (mod 4).
= (1)
p1 2
Theorem 5.
a1 a2 ...an p
a1 p
...
an p
Theorem 6 (Gauss criteria). For all a = 0 (mod p), p > 2, the following equality holds a p where l = |{ak | 1 k Theorem 7.
p1 , 2
= (1)l ,
p+1 }|. 2
ak (mod p)
p2 1 8
2 p
= (1)
+1, p = 8k 1, 1, p = 8k 3.
Theorem 8 (The quadratic reciprocity law). For all odd primes p = q the following equality holds: p q q p = (1) 17
p1 q 1 2 2
10 REPRESENTATIONS
Denition. Let odd m = p1 p2 ...ps , where pi are prime number, not necessary distinct, (a, m) = 1. a a = pa1 ... pas , where p are Legendres symbols. Then Jacobis symbols m i Theorem 1. The same as Theorem 1 for Legendres symbol. Theorem 2. The same as Theorem 2 for Legendres symbol. Theorem 4. 1 m +1, m = 1 (mod 4), 1, m = 3 (mod 4).
= (1)
m1 2
Theorem 5.
a1 ...as m 2 m
a1 as ... . m m +1, m = 8k 1, 1, m = 8k 3.
Theorem 7.
= (1)
m2 1 8
Theorem 8 (The reciprocity law for Jacobis symbols). Let m, n be odd numbers, m, n > 1, then n m
n1 m1 m = (1) 2 2 . n
10
Representations
k=0
In base b: Every n N0 can be uniquely written in base b, meaning n = ak {0, 1, 2, ..., b 1} and all but netely many ak = 0.
ak bk with all
Zeckendorf s (base Fibonacci) representation: Every n N can be uniquely expressed as a sum of Fibonacci numbers no two of which are consecutive. Warings Theorem: 18
12 PRIMES
Let f : N0 N0 by a polynomial and let d = gcd(f (0), f (1), f (2), ...). Then every sucient large multiple of d can be expressed as sum of a bounded number of values of f , or in other words: there is a k only depending on f such that for any n > N (N some constant) there are a1 , a2 , ..., ak N0 with dn = f (a1 ) + f (a2 ) + ... + f (ak ). Especially when 0 and 1 are in the range of f , then every n N0 can be written as a bounded number of values of f . Addionally, for any m there is a k such that any n N0 is the sum of k non-negative m-th powers of integers. Related to Warings Theorem: - every positive integer is the sum of 4 perfect squares (see also the Sum of Squares section). - every positive integer is the sum of 3 triangular numbers (those of type n(n2+1) ). - every integer is the sum of 9 non-negative perfect cubes. - every integer is the sum of 5 perfect cubes (they are allowed to be negative). Its an openen problem if 4 cubes suce.
11
p-adic valuations
Let p be any xed prime for this section. Properties of vp : For all rational a, b: vp (ab) = vp (a) + vp (b) Non-archimead triangle inequality: vp (a + b) min(vp (a), vp (b)) Hensels lemma: ps ||a 1, pk ||b, s 1 ps+k ||ab 1, or in other words vp (ab 1) = vp (a 1) + vp (b) for vp (a 1) 1, with exception of the case p = 2, s = 1. Kummers theorem: n If ps | n then s does not exceed the number of carries needed when the numbers k n k and k are added when expressed in base p.
12
Primes
Bertrands postulate 19
13 ADDITIVE PROPERTIES
There is always a prime between n and 2n (n N). Chebychevs Theorem: There are constants a, b, 0 < a < b such that for all big n (e.g. a = log(2), b = log(4) for n 2) we have a n (n) log(n) b n Prime number theorem x primes x. There are asymptotically log( x) Dirichlets theorem on primes in arithmetic progression: In every arithmetic progression an + b with gcd(a, b) = 1 there are innitely many primes. More exactly, the asymptotic and Dirichlets densities of these primes in the set of all primes are (1n) . Zsigmondys Theorem: Let a > b 1 and be coprime integers. Then for any n N there is a prime p dividing an bn but not dividing ak bk for all k < n with two exceptions: a) a = 2, b = 1, n = 6 b) a + b a power of 2 and n = 2
13
Additive properties
The Theorem of Chevalley-Warning: Let be p prime and f1 , f2 , ..., fm be m polynomials with integer coecients in the n variables x = (x1 , x2 , ..., xn ). If m i=1 deg(fi ) < n, then the number of solutions of f1 (x) 0 f2 (x) 0 ... fm (x) 0 mod p mod p mod p
is divisible by p (this generalizes to any nite eld). The Cauchy-Davenport Theorem: Let p be prime and A, B Z/pZ. Then the following inequality holds for the sumset A + B: |A + B | min(p, |A| + |B | 1) 20
14 MULTIPLICATIVE FUNCTIONS
Vospers Theorem (the case of equality in the Cauchy-Davenport Theorem): With the conditions above and A + B = Z/pZ, we have |A + B | = |A| + |B | 1 if and only if one of the following is true: a) |A| = 1 or |B | = 1 b) |A + B | = p 1 and B = (Z/pZ)\(c A), where c is the only one residue class A+B c) A and B are (seen mod p) arithmetic progressions with the same common dierence Some results that follow from the above: The Erd os-Ginzburg-Ziv Theorem: Let n N and 2n 1 integers be given. Then we can choose exactly n of them such that their sum is divisible by n. Sums of k -th powers mod p: Let p be prime and k N. Then mod p any number is the sum of k k -th powers, or in other words: for any n Z, there are integers a1 , a2 , a3 , ..., ak with k k k n ak 1 + a2 + a3 + ... + ak mod p. Sharper version of the previous one: 1 and k |p 1 (its With the same conditions as before, extended by p 5, 1 < k < p 2 clear that the condition k |p 1 is no restriction), and any n N we have that there 1 + 1 residues that are the sum of n k -th powers. are at least min p, (2n 1) p k
14
Multiplicative functions
21
16 FINITE DIFFERENCES
15
Irreducibility of polynomials
Theorem (Eisenstein) Suppose we have the following polynomial with integer coecients: f (x) = an xn + + a1 x + a0 . If there exists a prime p such that p|aj , j {0, 1, 2, . . . , n 1}, p an and p2 a0 , then f (x) is irreducible.
16
Finite dierences
Formula for n (f ):
n
f (x) =
r=0
(1)nr
n f (x + r) r
Eect on degrees of polynomials: When P is a polynomial of degree n, then k (P ) is a polynomial of degree n k , where negative degrees mean the constant polynomial 0 everytime.
22